Image segmentation is a method of subdividing an image into numerous meaningful regions or objects, which makes the image more informative and easy to analyze. Thresholding-based approaches are extensively employed for segmenting the image due to their low computational cost and are easy implementation. However, histogram-based thresholding schemes suffer from high variation that leads to abnormalities and sharp specifics. In this paper, we propose a technique for multilevel color image segmentation through variational mode decomposition (VMD) and Kapur’s entropy. Initially, the VMD is employed in order to decompose the histogram into corresponding submodes of analysis and attributes extraction, which leads to the removal of the unfavorable effects. Then, Kapur’s entropy is incorporated in order to generate accurate and optimal thresholds for segmentation. For the performance evaluation of the presented VMD–Kapur algorithm, various qualitative metrics have been used such as probability rand index, mean-square error, peak signal-to-noise ratio, variation of information, structural similarity index, dice error, feature similarity index, entropy, Jaccard/Tanimoto error, and normalized absolute error. The experimental results show that the proposed technique produces best-segmented images compared to Kapur’s, Tsallis, Masi, and fuzzy entropies.

中文翻译：

---

基于变分模式分解的彩色图像分割多级阈值选择方案

图像分割是一种将图像细分为许多有意义的区域或对象的方法，这使得图像信息量更大且易于分析。由于其计算成本低且易于实现，基于阈值的方法被广泛用于分割图像。然而，基于直方图的阈值方案存在高变异性，导致异常和尖锐的细节。在本文中，我们提出了一种通过变分模式分解 (VMD) 和 Kapur 熵进行多级彩色图像分割的技术。最初，VMD 用于将直方图分解为相应的分析和属性提取子模式，从而消除不利影响。然后，结合 Kapur 的熵以生成准确和最佳的分割阈值。对于提出的 VMD-Kapur 算法的性能评估，使用了各种定性指标，例如概率兰特指数、均方误差、峰值信噪比、信息变化、结构相似性指数、骰子误差、特征相似性指数、熵、Jaccard/Tanimoto 误差和归一化绝对误差。实验结果表明，与 Kapur、Tsallis、Masi 和模糊熵相比，所提出的技术产生了最佳分割的图像。骰子误差、特征相似度指数、熵、Jaccard/Tanimoto 误差和归一化绝对误差。实验结果表明，与 Kapur、Tsallis、Masi 和模糊熵相比，所提出的技术产生了最佳分割的图像。骰子误差、特征相似度指数、熵、Jaccard/Tanimoto 误差和归一化绝对误差。实验结果表明，与 Kapur、Tsallis、Masi 和模糊熵相比，所提出的技术产生了最佳分割的图像。